Research map / Video / Video generation
Improving Motion in Image-to-Video Models via Adaptive Low-Pass Guidance
June Suk Choi, Kyungmin Lee, Sihyun Yu, Yisol Choi, Jinwoo Shin, Kimin Lee
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| condition on image instead of text only (T2V to I2V) | Wan 2.1 · VBench prompts | Dynamic Degree: 39.4 -> 32.1 (-18.6%) | Hurt |
| condition on image instead of text only (T2V to I2V) | Wan 2.1 · VBench prompts | Imaging Quality: 68.3 -> 68.4 | No clear effect |
| apply stronger low-pass filter to the input image (ablation) | I2V model · VBench | Dynamic Degree | Mixed |
| add adaptive low-pass guidance (ALG) instead of CFG | Wan 2.2 · VBench-I2V | Dynamic Degree: 31.7 -> 39.0 | Improved |
| add ALG instead of CFG | Wan 2.1 · VBench-I2V | Dynamic Degree: 28.9 -> 39.4 | Improved |
| add ALG instead of CFG | LTX-Video · VBench-I2V | Dynamic Degree: 15.5 -> 21.5 | Improved |
| add ALG instead of CFG | Wan 2.1 · VBench-I2V | VBench-QS: 85.3 -> 84.5 | Hurt |
| add ALG instead of CFG | Wan 2.2 · PVD | Dynamic Degree: 65.0 -> 69.0 | Improved |
| add ALG instead of CFG | Wan 2.2 · VidProM | Dynamic Degree: 27.3 -> 30.5 | Improved |
| add motion-augmented prompts to ALG | ALG · VBench-I2V | Dynamic Degree: 39.0 -> 43.5 | Improved |
| increase transition time t_trans from 0 to 0.06 (ablation) | ALG · VBench-I2V | Dynamic Degree: increases 32% | Improved |
| add ALG (extra forward pass) instead of CFG | Wan 2.1 · Runtime | seconds: 476 -> 527 | Hurt |
Closest papers by meaning
- Extrapolating and Decoupling Image-to-Video Generation Models: Motion Modeling is Easier Than You Think (CVPR 2025 · similarity 0.89)
- Rebalancing Reference Frame Dominance to Improve Motion in Image-to-Video Models (NeurIPS 2026 · similarity 0.89)
- Encapsulated Composition of Text-to-Image and Text-to-Video Models for High-Quality Video Synthesis (CVPR 2025 · similarity 0.89)
- VideoGuide: Improving Video Diffusion Models without Training Through a Teacher's Guide (CVPR 2025 · similarity 0.88)
- SG-I2V: Self-Guided Trajectory Control in Image-to-Video Generation (ICLR 2025 · similarity 0.87)
- Spatiotemporal Skip Guidance for Enhanced Video Diffusion Sampling (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
